Efficient 2-Segment Routing Based on Local Search With Failure Recovery

Jingjing Chen, Xueyun Ling, Ying Wang, Xuesong Qiu

Abstract: Segment Routing (SR) is a flexible and efficient source-routing technology. It can forward traffic along arbitrary paths in the network and has good scalability without the maintenance of routing information at intermediate nodes. These advantages make SR widely used in data centers, WANs, MANs, and other networks, particularly in the context of parallel and distributed processing. However, current SR traffic engineering has some shortcomings in terms of network load balancing, failure resiliency, and complete use of SR characteristics, which prevents better optimization of comprehensive network performance. In this paper, we propose a heuristic traffic engineering algorithm (2-SRLS) based on a two-segment routing model (2-SR), which incorporates a traffic-splitting strategy, adjacency segments and failure resiliency. The algorithm supports the technical characteristic of adjacency segments in SR with a flexible source node traffic splitting strategy and can efficiently recover from single-link failures. Experimental results show that our algorithm is close to the theoretical optimum in terms of the performance of reducing the maximum link utilization and has a 10%-16% improvement in active link coverage compared to existing algorithms. The running time is reduced by 70% compared to existing heuristics. In addition, the proposed algorithm can effectively recover from single-link failures on the basis of guaranteed the stability of the maximum link utilization (MLU).